Journal of Hydrology 517 (2014) 284—297

IJOURNAL OF

HYORDLD6Y

1. Introduction

Urban growth in the arid and semi-arid regions of the UnitedStates and other countries places significant stress on waterresources, which in many localities are already stressed due to limited recharge and increased water demand. While characterizationof water resources is always desirable, accurate assessment ofwater availability in areas where resources are limited and stressedis of critical importance. Due to the unique and complex groundwater hydraulics of carbonate aquifers, special considerations arewarranted when characterizing and managing water resources insemi-arid environments. Carbonate aquifers can serve as the principle source of water in a semi-arid environment as occurs in Spain(Hartmann et al., 2013; Martinez-Santos and Andreu, 2010),Lebanon (Bakalowicz, 2005; El-Hakim and Bakalowicz. 2007) andTexas, USA (Anaya and jones, 2004, 2009; Hutchison et al., 2011;Green and Bertetti, 2010) and for this reason, accurate characterization of the aquifer system is paramount.

Understanding the means and mechanisms by which carbonateaquifers convey water from the headwaters of the watersheds totheir points of discharge is important to the effective management

* corresponding author. Tel.: +1 210 522 5305; fax: +1 210 522 5184.E-mail address: rgreen@swri.edu (R.T. Green).

of these valuable resources. The degree of karstification determineswhether groundwater flow can be characterized as Darcian or isdominated by conduit flow (Scanlon et al., 2003; Worthington,2007; Rashed, 2012). Conduit flow can be detected directly withdye tracer tests and indirectly using other hydraulic factors, suchas groundwater gradients (i.e.. troughs) and aquifer response (i.e.,spring discharge) (Schindel et al., 1996; Worthington et al., 2000;Worthington, 2007). Rarely, however, are sufficient site-specificdata available to adequately characterize the hydraulic propertiesof a karst-dominated aquifer to allow for effective managementof the resource.

Characterizing karst-dominated aquifers that exhibit well-developed preferential flow paths and permeability architecturesspanning many orders of magnitude can be challenging. Practitioners have used various tools to aid in characterizing preferential flowpaths in karst systems. Considerable effort has been expended to uselineaments and topographic expressions to discern subsurfacehydraulic properties (Lattman and Parizek, 1964; Parizek, 1975;Sander et al., 1996; Magowe, 1999; Mabee et al., 1994, 2002;Moore et al., 2002; Mouri, 2004; Bauer et al., 2005; Mouri andHalihan, 2007).

To characterize the preferential flowpaths of a karst-dominatedaquifer, a method is proposed that recognizes the importance oflineaments and topographic expressions, the principles of

ELSEVIER

Contents lists available at ScienceDirect

Journal of Hydrology

journal homepage: www.elsevier.com/locate/jhydrol

focused groundwater flow in a carbonate aquifer in a semi-aridICrossMark

environment

R.T. Green ‘, F.P. Bertetti, M.S. MillerGeosciences and Engineering Division, Southwest Research lnstitute, United States

ARTICLE INFO SUMMARY

Article history: An efficient conveyance system for groundwater is shown to have formed in a carbonate aquifer evenReceived 21 October 2013 though it is situated in a semi-arid environment. This conveyance system comprises preferential flowReceived in revised form 4 April 2014 pathways that developed coincident with river channels. A strong correlation between high capacityAccepted 3 May 2014 wells and proximity to higher-order river channels (i.e., within 2.5 km) is used as evidence of preferentialAvailable online 20 May 2014

flow pathways. Factors that contributed to development of the preferential flow paths: (i) karst develop-This manuscript was handled by Peter 1<.Kitanidis, Editor-in-chief, with the ment in carbonate rocks, (ii) structural exhumation of a carbonate plateau, and (iii) the requirement that

assistance of Barbara Mahler, Associate the groundwater regime of the watershed has adequate capacity to convey sufficient quantities of waterEditor, at the required rates across the full extent of the watershed. Recognition of these preferential pathways in

proximity to river channels provides a basis to locate where high capacity wells are likely (and unlikely)

Keywords: and indicates that groundwater flow within the watershed is relatively rapid, consistent with flow ratesl<arst hydrology representative of karstic aquifers. This understanding provides a basis for better informed decisionswater-budget analysis regarding water-resource management of a carbonate aquifer in a semi-arid environment.Groundwater conveyance © 2014 Elsevier By. All rights reserved.well capacityArid-land rechargeCarbonate aquifer

http://dx.doi.org/10.101 6/j.jhydrol.20 14.05.0150022-1694/© 2014 Elsevier By. All rights reserved.

R.T. Green et al/Journal of Hydrology 517 (2014) 284—297 285

carbonate dissolution, and a surrogate estimate of aquifer permeability. Spatial distribution of well capacity is used to establish acorrelation between preferential flow paths in the carbonate aquifer and proximity to river channels. Other spatial relationshipswere explored, such as correlations between well pumpingcapacity and geology, geomorphology, or karst features, however,only well pumping capacity and proximity to river channelsdemonstrated a useful correlation. The ensuing network of preferential flowpaths that are co-aligned with river channels is thefoundation for a refined conceptual model in which flow in thewatershed is dominated by the preferential flow paths ratherthan diffuse flow through the inter-stream upland areas. Correlation between karst development and river channels has beenobserved elsewhere (Abbott, 1975; Woodruff and Abbott, 1979,1986; Allen et al., 1997; MacDonald and Allen, 2001;Mocochain et al., 2009), however the use of well hydraulics hasnot been used to quantify the degree of karst developmentaligned with river channels.

The Devils River watershed in south-central Texas, USA (Fig. 1) isselected to test this method because it conveys significant groundwater in a semi-arid environment and because it is representativeof a broader class of carbonate aquifers in semi-arid environmentsworldwide. Accordingly, characterizing key groundwater conveyance mechanisms in the Devils River watershed may help characterize similar karst aquifers in other semi-arid environments.

2. Geological and hydrogeological setting of the study area

The carbonate aquifers in central Texas, USA are the primarysources of water for a rapidly growing population. Most prominentof these are the Edwards, Trinity, and the Edwards-Trinity aquifers.These aquifers exhibit a broad range of hydraulic characteristics. Ofinterest is the western Edwards-Trinity Aquifer, an exhumedcarbonate aquifer which is the source for significant water

resources, although it is located in a semi-arid environment. TheDevils River watershed, located in the western Edwards-TrinityAquifer (Fig. 1), exhibits aquifer and hydraulic characteristics representative of the greater Edwards-Trinity Aquifer and parts of theTrinity Aquifer, but distinct from the Edwards Aquifer (Abbott,1975; Woodruff and Abbott, 1979, 1986).

The Edwards-Trinity Aquifer covers 200,000 km2 and is thedominant aquifer in west-central Texas (Barker and Ardis, 1996)(Fig. 1). This Cretaceous-age limestone comprises the younger,more permeable Edwards Group rocks overlying the older and lesspermeable Trinity Group (Fig. 2). The Edwards-Trinity Aquifer hassignificant vertical and lateral spatial variability (Rose, 1972). Theclimate varies from humid subtropical in the east to arid andsemi-arid (steppe) in the west. The Devils River watershed conveysan average of 324 Mm3/yr of water from the Edwards Plateau to theAmistad Reservoir and the Rio Grande in the south. This amounts toover 15% of the total flow of the lower Rio Grande (United StatesGeological Survey, 2013)—an impressive quantity of water delivered from a semi-arid area where average precipitation is less than500 mm/yr over a surface watershed comprising 10,260 km2.

Most of the Edwards Plateau is mantled by the Edwards Formation with a tableland geomorphological surface that exhibits astair-step topography formed by differential weathering of stratawith variable resistance. More resistant layers form treads” whichare gently sloping surfaces with minimal (i.e., <0.5 m) soil overburden. Less resistant layers weather to form “risers”, step-likefeatures with clay-rich, low-permeable soils with a thickness ofless than 1.0 m to as much as 3.0 m (Woodruff and Wilding,2008; Wilcox et al., 2007). The Devils River is incised through thetableland surface exposing steep cliffs in places. Aside from theincised river and stream channels there are few karst features suchas sinkholes or other solution cavities exposed at the surface.

Geologic mapping is useful in characterizing the hydraulicproperties of an aquifer when site-specific studies have not been

Fig. 1. Location map of the Devils River basin and the Sycamore creek sub basin in central Texas.

286 RE Green et aL/Journal of Hydrolo’ 517(2014)284—297

Geo- Western EasternAquifer

Chronology N Study Area S N Study Area S

Quatemary Alluvium Alluvium

I Uvalde Gravel

)Late

Tertiary

Del Rio Rio —

OoquilasfEagle FordCretaceous

stin

Buda

SalmonFort Peak Segovia

Lancaster 0,5, 5> ‘ °‘

4: <McKnight c C,,

ECoEarly

FortCretaceous Fort ‘o LuLu TerrettTerrett

Nueces :

—.

C,),

Antlers San-_

Lu

‘.-

Basaleou’ •5 V

sell>‘M o>a’

SandSand :

jcor= -

Hosslon

Copper CanyonEILate I Trujio Undivided o

—01

TecovasTriassicSanta Rosa

Permian

Ordovician

Cambrian

Undivided

Undivided

Undivided

Ellenburger

San Saba

0 C

Undivided

Fig. 2. Stratigraphic column and major aquifer units for the Devils River region and Edwards Plateau.

performed and aquifer characterization data are not available. Conventional characterization of the hydraulic properties of theEdwards-Trinity Aquifer in the Devils River watershed basin hasbeen based on its mapped geology (Anaya and Jones, 2004, 2009;Hutchison et al., 2011). This characterization is well illustratedby the hydraulic conductivity assigned to the current groundwaterflow model used to manage the Edwards-Trinity Aquifer (Fig. 3)(Hutchison et al., 2011). Although most hydraulic property assignments are consistent with the mapped geology in Fig. 3, someassignments of the hydraulic conductivity values are ambiguous(Table 1).

Obviously, supplemental hydrogeological information canprovide additional insight when characterizing an aquifer than isprovided by geologic mapping alone. This is the case with theEdwards-Trinity Aquifer in the Devils River watershed basin. Therecognition that the Edwards-Trinity Aquifer is a karst aquifer, inwhich preferential pathways have developed in the limestone, isparamount. In this case, assigning hydraulic properties to a karstaquifer based solely on geologic maps does not take intoconsideration the dominating effect of preferential flow pathspresent in the Edwards-Trinity Aquifer.

3. Preferential flow path development

Refined hydraulic properties are proposed for the Devils Riverwatershed basin based on data and information now available thatprovide insight regarding preferential pathways in the Edwards-Trinity Aquifer. The interpretation developed in this paper is thatpreferential pathways have developed coincident with river channels in the Edwards-Trinity Aquifer and that these preferentialpathways are the principal means of conveying groundwater fromthe watershed’s headwaters to its points of discharge, The development of organized flow regimes forming in karst systems is notuncommon (Bakalowicz, 2005; El-Hakim and Bakalowicz, 2007;Worthington and Ford, 2009).

Factors that controlled conduit development in the Devils Riverwatershed were (i) the degree to which rocks are susceptible todissolution, (ii) exhumation of the Edwards Plateau leading toincreasing the effective hydraulic gradient, and (iii) the focus ofrecharge into a defined stable river system (White and White,2001). Palmer (1991) notes that cave patterns with limitedbranches tend to form if recharge is focused, the carbonate rockis limestone rich, and hydraulic gradients are at least moderate

RT. Green et al/Journal of Hydrology 517 (2014) 284—297 287

Fig. 3. Map of Devils River watershed illustrating (a) geologic assignments based on State of Texas geologic maps (Fisher, 1977, 1981) (left) and (b) hydraulic conductivityvalues assignments (taken from Hutchison et al., 2011) (right). The red border denotes the Devils River watershed basin.

Table 1Assignment of hydraulic conductivity values to Devils River basin rocks based on geologic mapping (extracted from Anaya and Jones, 2004, 2009; Hutchison et al., 2011).

Hydraulic conductivity (rn/day) Geologic formation Geographical feature

17—24 Buda Limestone, Kbu Edwards Plateau0—1.2 Segovia, Ks Southern end of Edwards Plateau1.5—4.2 Segovia/Buda Urnestone, Ks/l<bu Eastern Devils River Channel1.5—4.2 Segovia, Ks North-central Devils River Channel4.6—8.8 Del Rio Clay, l(dr South-central Devils River Channel9.1 —16.5 Salmon Peak, Nsa Southeast Devils River Channel0—1.2 Del Rio Clay/Buda Limestone/Eagle Ford, l<dr/Kbu/Kef South Devils River Channel.

(i.e.,>0.001). White and White (2001) concur that hydraulic gradients of 0.001 and greater are adequate to enable the developmentof focused conduits. Alternatively, low hydraulic gradients couldhave led to multiple alternate flow paths.

Exhumation of the Edwards Plateau and the subsequentdevelopment of the groundwater conveyance system in theEdwards-Trinity Aquifer occurred during two diverse episodes.The first episode was in the middle Cretaceous Period when theEdwards Group limestones were deposited, subaerially exposed,and then buried. The second episode started during the MioceneEpoch when Balcones faulting eroded the fault-rejuvenatedstreams and exhumed the Edwards Group limestones (Abbott,1975; Woodruff and Abbott, 1979, 1986) and is potentiallyongoing. Exhumation of the karstic tablelands preserved relictlandforms such that streambeds became incised valleys whoseevolution was enhanced by increased hydraulic gradients.

During uplift, incipient preferential flow paths formed in thesubsurface, coincident with the existing river systems, when

mildly acidic precipitation flowed in riverbeds and developedenhanced permeable flow channels in the soluble carbonate rock.The susceptibility of limestone to dissolution is a function of theamount of calcium carbonate in the rock (Dreybrodt andGabrovsek, 2003; Romanov et al., 2003; Dreybrodt et a!., 2005).Solution features, such as conduits and other karst features, developed in carbonate rocks when weak carbonic acid formed fromrainwater and organic carbon dissolved calcium carbonate (CaCO3)over geologic time (Ford and Williams, 1989)

CaCO3 -- CO2 — H20 = Ca2 2HC03

Aquifers with greater limestone content tend to have betterdeveloped conduit systems, resulting in primarily conduit flow inthe aquifer. Conversely, aquifers with higher dolomite fCaMg(C03)2] content tend to be less susceptible to carbonate dissolutionthan aquifers with greater limestone content, although karst development is certainly observed in dolomite-rich formations (Whiteand White, 2001).

(a) (b)

288 R.T. Green et at/Journal of HydroIoi 517 (2014) 284—297

Geologic lineaments and zones of fracture concentration havebeen shown to act as avenues for enhanced weathering andincreased permeability, thereby facilitating vertical and lateralgroundwater movement (Siddiqui and Parizek, 1971; Parizek,1975; Lattman and Parizek, 1964; Sharpe and Parizek, 1979;Klimchouk and Ford, 2000a,b). Once initiated, the preferential flowpaths were further enhanced by a positive-feedback growth mechanism in that an increased volume of mildly acidic water was available to promote solution cavity development. This preferentialflow-field development converged in river channels because thetopography channeled water from uplands to the river channelswhere dissolution was concentrated in the shallow phreatic zone(Abbott, 1975) (Fig. 4). Similar genesis of enhanced permeabilitynear river channels has been observed in the unconfined ChalkAquifer in England (Allen et al., 1997; MacDonald and Allen, 2001).

Uplift and contemporaneous faulting at the boundary of theEdwards Plateau increased hydraulic gradients that incised intothe limestone plateau. The incised valleys often led to topographical low points, providing for spring discharge. Watershed piracyfrom cut-off streamfiow and fault-induced watershed interconnection in the eastern Edwards Aquifer allowed for more direct surfaceflow paths with increased hydraulic gradients (Woodruff, 1974,1977; Woodruff and Abbott, 1986). Because the same conditionsexisted south of the Edwards Plateau that existed in the easternEdwards Aquifer, similar evolution of surface-water flow regimesin the Devils River watershed basin would also have led toincreased hydraulic gradients.

Using the potentiometric surface of the Edwards-Trinity Aquifer(Kunianslcy and Holligan, 1994; Barker and Ardis. 1992, 1996; Bushet al., 1993; Ardis and Barker, 1993), current hydraulic gradientshave been measured in proximity to the Devils River watershedbasin. The gradients are 0.0016 in Sutton County, 0.0013 in ReaganCounty, 0.0012 in Crockett County, and 0.0038 in Val Verde County.These measured hydraulic gradients are sufficiently large to support the development of preferential flow paths focused in riverchannels rather than expansive solutionally enhanced flow spreadover broad paths.

There is evidence that another form of piracy, in which groundwater basins extend farther upgradient than the overlying surfacewatersheds, exists in the western Edwards Aquifer (Woodruff andAbbott, 1979, 1986; Green and Bertetti, 2010). The resultingenhanced flow regime, whether due to a longer flow path or toan increased hydraulic gradient, increases the degree of positivefeedback in the development of solution features in the karsticlimestone. This in turn leads to further development of the karsticflow regime and enlargement of lower level conduits at the pointsof discharge (Woodruff and Abbott, 1986). The hypothesisproposed here is that a number of smaller conduits have formed

proximal to river channels rather than one large conduit or evena small number of larger conduits. The ensemble of the smallerconduits has resulted in high preferential flow paths formedconcurrent with river channels.

The variable susceptibility of carbonate dissolution in the formations of the Edwards Plateau has important implications withregard to how conduits form in the Edwards-Trinity Aquifer. TheUpper Glen Rose member of the Trinity Group has been categorized as predominantly a thin- to medium-bedded sequence ofnonresistant marl alternating with resistant beds of dolostone,lime mudstone, and bioclastic limestone (Stricklin et al., 1971;Barker et al., 1994). Lower units in the Trinity Group also tend tobe less rich in limestone relative to the Edwards Formation andto the Upper Glen Rose member; thus dissolution of the carbonaterocks in the Edwards Formation is likely to be more rapid or moredeveloped relative to the Trinity Group. In brief, carbonatedissolution is going to occur preferentially in the limestone-richEdwards Limestone portions of the Edwards-Trinity Aquifer rocksrelative to the less limestone-rich and low-permeability rockspresent as interbeds in the Trinity Group.

The upper Edwards Group limestone (i.e., Segovia Member) ismostly exposed at the surface in the Edwards Plateau, an area thatincludes the Devils River watershed (Fig. 3). There are limitedoccurrences of the overlying Buda Limestone, in which cases thefull thickness of the Edwards Group is preserved. Elsewhere, erosion has removed all formations that overly the Edwards Groupand part of the upper Edwards Formation leaving only a variablethickness of the Edwards Group present. Although the upperEdwards Group has been eroded over most of the Edwards Plateau,at no place in the Devils River channel has the Edwards Group beenfully eroded to expose the Trinity Group (Fisher, 1977, 1981). Thisfactor is important because the limestone-rich Edwards Grouplimestone is available to provide for conduit developmentthroughout the entire reach of the Devils River channel.

4. River channel groundwater flow regime development

Flow in the Devils River has been synoptically measured twiceand has had continuous measurements taken at two locations atvarious times during the past 50 years. The Cauthon Ranch rivergauge located near Juno is in the upper reach near the headwaters.The Pafford Crossing gauge location is located at the Devils lowerreach, slightly upstream from where surface water has backed upin the Devils River since the Rio Grande was dammed in 1969, creating the Amistad Reservoir. Table 2 lists average flow versusdrainage area for the two gauge locations for two different periodsof record.

NUECES RIVER BASIN GUADALUPE RIVER BASIN

Erosion of overlying rockRelict coves in vodose zone;high korStic plains

Maximum discharge nearSites

Long- distance grounth watertronstor otong strike

Fig. 4. Schematic cross section of the development of recharge caverns coincident with incised river channels in the Western Edwards Aquifer (taken from Woodruff andAbbott, 1979, 1986).

RI. Green ec al/Journal of Hydrology 517 (2014) 284—297 289

Table 2Average flow (L/min and Mm3/yr) measured at two gauging locations on Devils River.The Cauthon Ranch nearJuno values are the averages of annual measurements for theperiods of 1926—1949 and 1964—1973. The Pafford Crossing values are the averages ofdata of daily measurements for the period 1/1/1960 to 12/31/2011.

Gauging station Drainage area (km2) Flow (Mm3/yr) flow (Mm3/yr)

Cauthon Ranch 7164 168 149Pafford Crossing 10,256 323 294

Over both time periods, river flow increased by over 90% (92.6%in column 4 and 96.5% in column 5) between the Cauthon Ranchriver gauge and the Pafford Crossing rivet gauge, even though thedrainage area only increased by 43%. The obvious source ofincreased flow between these two gauging stations is due to emergent flow in the river channel, not to the incremental increase inthe size of the watershed between the Cauthon Ranch and PaffordCrossing river gauges. This observation is consistent with a conceptual model of preferential flow path development in river channelsand that the preferential flow is increasingly discharged to theriver in downstream reaches.

Synoptic flow surveys, also referred to as gain/loss surveys, havebeen completed twice on the Devils River. Synoptic flow surveysare valuable for several reasons, one of which is to identify whichriver reaches are gaining and which are losing. A synoptic surveyconducted under low flow conditions on the Devils River was completed in July 2013 and compared with a published synoptic surveyconducted under relatively higher flow conditions during anunspecified time in 2006 (Texas Commission for EnvironmentalQuality, 2006) (Fig. 5). With the exception of two minor decreasesin measured flow made in 2006 in the upper reach and one readingof 1198 L/s in the upper reach during the 2013 survey, the entirereach of the Devils River was gaining from its headwaters to itsoutfall into the Amistad Reservoir. This is particularly obviousdownstream from the confluence of Dolan Creek with Devils River.In addition, the river gains at a rate in excess of the increase inwatershed area. This excess in increased flow is attributed tocontributions from subsurface channel flow mostly attributed todiscrete discharge at spring locations.

Visual inspection of the Devils River indicates the river bed ismostly exposed bedrock with minimal evidence of gravels or otherfloodplain sediments. This observation supports the hypothesisthat the increase in surface flow in the Devils River is attributableto contributions from bedrock and not from hyporheic flowthrough gravel and other riverbed sediments.

5. Well capacity correlated with river channel proximity

The principle hypothesis proposed in this paper is that preferential flow paths and enhanced permeability are coincident withriver channels in the Devils River watershed. A similar correlation,in terms of high transmissivity aligned with River Lambourn andRiver Kennett, was discerned in the unconfined Chalk Aquifer inBerkshire County, England (Connorton and Reed, 1978; Morel,1980; Allen et al., 1997; MacDonald and Allen, 2001). The hightransmissivityfriver correlation of the Chalk Aquifer was discernedusing geomorphological evidence. The hypothesis is tested on theDevils River watershed by correlating water-well pumpingcapacity and well proximity to river channels. Well data from theDevils River watershed and an adjoining minor watershed, theSycamore Creek watershed, were extracted from the Texas WaterDevelopment Board Submitted Driller’s Report and Groundwaterdatabases (VWDB, 2012a,b) to assess whether the hypothesis ofthe development of preferential flow paths and enhanced

permeability coincident with river channels in carbonate aquifershas merit. These databases are the most comprehensive datasetsavailable for the Sycamore Creek and Devils River watersheds thatprovide some measure of hydraulic capacity. Unfortunately,hydraulic capacity information is limited to well pumping capacity.A more useful measure, such as specific capacity, is not included ineither database. Information on pumping capacity is included for751 of the 2122 wells in the database for the Sycamore Creekand Devils River watersheds. The remaining wells have either norecord of pumping capacity or limited pumping capacity. Domesticor stock wells with limited pumping capacity (i.e., less than 75 L/mm) are believed to comprise the bulk of the wells with no recordof pumping capacity. Limited field checking failed to identify anyadditional wells with significant pumping capacities that werenot included in the subset of 751 wells. Locations of the wells withmeasured pumping capacity in the Devils River and SycamoreCreek watersheds are plotted in Fig. 6.

Proper selection of river channels is critical to the correlation ofwell pumping capacity with proximity to stream channels. Watersheds, such as that of the Devils River, with low annual rainfalltotals, high intensity rains, and sparse vegetation have high drainage density (Gregory, 1976; Rodriguez-Iturbe and Escobar, 1982).Given the high density of incised and intermittent stream valleyswithin the Edwards Plateau, the correlation of well pumpingcapacity with proximity to river channels would be biased ifstream channels were fortuitously selected to use only those channels proximal to each well with high pumping capacity. Onlystream segments with a Horton—Strahler number of three orgreater as classified in the National Hydrography Dataset, Version2 (United States Geological Survey. 2013) were included in thisanalysis to avoid selection bias.

The ArcGlS geoprocessing tool Near was used to calculate distances between wells and third-order and greater streams. Thiscomputation was facilitated by entering shape files for the third-order and greater streams (United States Geological Survey,2013) and locations of all wells that had documented values forwell capacity. Each well with a documented pumping capacitywas thereby assigned an unambiguous measurement that represented the shortest distance to the closest third-order stream.Wells with no pumping capacity measurement were excludedfrom the evaluation.

The correlation between well pumping capacity and proximityto stream channels is presented in Fig. 7. A strong correlationbetween well capacity and proximity to higher-order river channels is clearly illustrated in the graph. Care must be taken wheninterpreting well pumping capacity. Although the measuredpumping capacity of a well may represent its maximum capacity,it is probably less than the potential maximum pumping capacityof the well at its location. It is possible that a large, possibly deeper,well with a larger pump at the well’s location would have greaterpumping capacity. Regardless, it is significant that out of a datasetof 2122 wells of which 751 wells were assigned a value for pumping capacity, only one well with pumping capacity greater than1890 L/min is located more than 2.5 km from third-order or largerstreams.

Because of the uncertainties associated with the well pumpingcapacities in the Driller’s Report database, it is fruitful to analyzethe relationships between the total number of wells with availablepumping data and distances to the neatest higher-order river channel. A histogram of the well pumping capacity data in terms oflog10(L/min) is shown in Fig. 8. The data are adequately represented by a log-normal distribution with a mean (rl) of 1.83(67 L/min) and standard deviation (u) ± 0.65. Four classes of wellpumping capacity were defined using the distribution parametersp and u: low p (67 L/min); med tl + 0 (68—300 L/min);p + a> med-high p + 2u (301—1300 C/mm); and high> p + 2u

290 RI. Green et at/Journal of Hydrolo’ 517 (2014) 284—297

(>1301 C/mm). The well pumping classes were then binned in1-1cm increments with respect to their calculated distance to thenearest higher-order river channel. A bar graph of the results isshown in Fig. 9. Data in the figure demonstrate that both wellpumping capacity (i.e., frequency of greater capacity well classes)and the total number of wells increase dramatically as the distanceto the nearest river channel decreases. A majority of the wells fallinto the 67 C/mm classification, but these wells are also locatedwithin a few 1cm of the higher-order stream channels. Fig. 10 presents the same data plotted on a 1og10-log1o scale. The data are separated into two groups, one that includes all the analyzed wellsand a second that excludes the 67 L/min pumping capacity wells.Both data sets are well characterized by a power function fit of theform Count = a.(Distance)b, where a and b are constants. The observation that distance from river channels is a reasonable predictorof well pumping capacity and well count is similar to trendsobserved in other studies of carbonate aquifer properties (e.g.,MacDonald and Allen, 2001). When combined with data in Fig. 7,these trends provide compelling evidence that wells with a highpumping capacity are restricted to areas in close proximity to riverchannels.

6. water-budget analysis

Having an understanding of the water budget of the Devils Riverwatershed can help constrain the groundwater/surface water flowconceptual model. Historically, the water budget of the DevilsRiver watershed basin has not been well characterized. Althoughdischarge from the Devils River to Amistad Reservoir is measured,uncertainty remains regarding the size of the Devils River rechargebasin and the rate of recharge within the basin. Accurate calculation of recharge is obviously central to meaningful determinationof the water budget in any environmental setting and carbonateaquifers in a semi-arid climate are no exception. A variety ofapproaches has been used to calculate recharge in semi-arid environments. Baseflow separation has been used to estimate rechargewhen spring or stream discharge measurements are available(Arnold et al., 1995; Arnold and Allen, 1999; Green and Bertetti,2012; Green et al., 2012). Chemical analysis of spring or streamdischarge has been shown to be useful to either assist in baseflowseparation (Doctor et al., 2006) or to discern seasonal contributionsto recharge and, ultimately, discharge (Aquilina et al., 2005).Another approach to determine recharge is to evaluate the spatial

Fig. 5. Locations of flow measurements along the Devils River during relatively high flow conditions in 2006 (TCEQ 2006) and low flow conditions in 2013.

RE Green et at/Journal of Hydrology 517 (2014) 284—297 291

and temporal variability of recharge using aquifer response(Hughes and Mansour, 2005; Hartmann et al., 2012, 2013).

Recharge to the Devils River watershed is relatively significantgiven the amount of water discharged by the Devils River to Amistad Reservoir even though the watershed is located in a semi-aridenvironment (Reeves and Small, 1973; Veni, 1996; Green andBertetti, 2010). Flow at the Devils River Pafford Crossing gaugelocated near Amistad Reservoir is typically referenced as the measure for average discharge from the Devils River to Amistad Reservoir. This discharge of 324 Mm3/yr accounts for approximately 15%of the flow in the lower Rio Grande (1973 Mm3/yr) (InternationalBoundary and Water Commission, 2005). Precipitation recharge

in counties that cover the Devils River watershed basin has beenapproximated at 7.9—12.4 mm/yr by Hutchison et al. (2011) usinga groundwater model (Table 3). Recharge in Val Verde County waspreviously estimated at 38.1 mm/yr by Reeves and Small (1973)and in the Dolan Creek tributary to the Devils River watershed inVal Verde County at 55.4 mm/yr by Veni (1996).

Recharge can also be established using precipitation measurements. Average annual precipitation from 1971 to 2000 for theDevils River watershed area is mapped in Fig. 11. As illustratedin Fig. 11, the average annual precipitation for each county withinthe Devils River watershed varies from less than 400 mm/yr in thewest to about 585 mm/yr in the east (Table 3). Recharge for the

Fig. 6. Map of the Devils River watershed with well locations. Highest capacity wells [>3785 C/mm (1000 gpm)] are denoted by a red dot, higher capacity wells Ibetween1890 and 3784 L/min (500 and 999 gpm)l are denoted by a yellow dot, lower capacity wells [between 378 and 1889 C/mm (100 and 499 gpm)] are denoted with green dots,and wells with capacity less than 100 gpm are denoted with a purple dot. As illustrated, the majority of wells have capacities less than 378 L/min (100 gpm).

292 RI. Green et al/Journal of Hydrology 517 (2014) 284—297

10000

8000

IsC.

6000soC

2000

fig. 7. Graph of well capacity versus distance to the nearest third-order or greaterstream channel. The data suggest a strong correlation between distance and wellpumping capacity, especially for higher capacity wells (>2000 C/mm).

,

0 1 2 3 4 5

log15 Well pumping capacity (L/mln)

Fig. 8. Histogram and fit of normal distribution for log1>-transformed well pumpingcapacity data for Devils River region wells. The data are reasonably represented by alog-normal distribution with log1op of 1.83 C/mm and log10a of ±0.65.

1 10 100

Distance (km)

Fig. 70. Scatter plot of classified well pumping capacity data versus distance to thenearest third-order or greater stream channel. The data indicate distance fromstream channels is a good predictor of well pumping capacity and the number ofwells. The lines represent least-squares best fits of power functions to each of thedata sets and are presented with their respective R2 values.

12000

C0

1000

100

10

10

.All Wall Punpng tlaa

OMed. MedHigh, and High

A HighRI=082

0.83p ‘..0

-.,

A

0

0’ = A

5 10 15 20

Distance to Nearest Stream Channel (km)

360

270

C

0U

ISO

90

00

04 Devils Rivet basin has been estimated based on an assessment ofthe western Edwards-Trinity Aquifer in which avetage rechargewas shown to correlate linearly with average precipitation, but

03 .g to become negligible when precipitation is less than 400—430 mm/yr (Green and Bertetti, 2010; Green et al., 2012). Thisthreshold is important when assessing recharge; however it must

02‘ be recognized that it is a long-term average and that factors such

as antecedent moisture, storm duration and intensity, and seasonal

0 factors such as temperature and humidity have a significant impacton actual recharge rates. Using the approach by Green and Bertetti(2010) and Green et al. (2012), average recharge is estimated torange from 16.0 to 33.0 mm/yr by), for an average annual rechargevalue for the Devils River 20 mm/yr. It is important to recognizethis rate has significant spatial and temporal variability given thefact that the watershed is near this calculated threshold for negligible distributed recharge and that average precipitation varies significantly across the watershed (Fig. 11).

C

U

250

200

150

lao

50

0

• High (>1301 lJmln)

U MedHigh (301-1300 1/mm)

Med (68-3C0 l./min)

El Low ( 67 C/mm)

1234 567891011121314151617181920212223Z4Z5

Distance (km)

fig. 9. Number of wells in each pumping capacity class plotted by distance to the nearest third-order or greater stream channel. Wells clas5es are based on mean andstandard deviation values for the log-normally distributed pumping capacity data for wells in the Devils River region.

RE Greener al/journal of Hydrology 517 (2014) 284—297 293

Table 3Comparison of recharge by Hutchison et a!. (2011) and Green and Rertetti (2010) for Counties within the Devils River watershed.

County Precipitation (mm/yr) Recharge (mm/yr) Recharge’ (mm/yr)

Range Average

Crockett 380—530 530 12.4 8.6Edwards 580-740 530 1 1.7 33.0Schleicher 530—580 560 7.9 20.0Sutton 530—610 530 10.2 25.4Val Verde 430—530 510 9.9 16.0

Baseflow and surface runoff were separated from flowmeasurements using the Devils River Pafford Crossing gaugedata collected during the period 1960—2009 (Arnold et al.,1995; Arnold and Allen, 1999). Baseflow was calculated to be76% of total flow with the remaining 24% contributed by

surface runoff (Green and Bertetti, 2010; Green et al., 2012).Thus, 76% of the 324 Mm3/yr (or 246 Mm3/yr) the Devils Riverdischarges to the Amistad Reservoir is attributed to baseflowand, hence, recharge (White and White, 2001: White, 1999,2006).

a Hutchison et al. (2011).Green and Bertetti (2010).Average precipitation within the Devils River watershed located within each county.

Fig. 11. Average annual precipitation (mm/yr) in the Devils River region over the period 1971—2000. Precipitation data are courtesy of PRISM Climate Group, Oregon Stateuniversity, http:•.nrisrn.oregonsrate.edu, created 06 December 2006.

294 R.T. Green et al/Journal of Hydrology 517 (2014) 284—297

If recharge for the Devils River watershed basin is estimated at11.4mm/yr (area! average of recharge for the Devils Riverwatershed basin estimated using countywide recharge values byHutchison et al., 2011), then 21.583 km2 of watershed is required

10,260 1cm2 and discharge rate of 246 Mm3/yr to the Amistad Reservoir. These calculations suggest that the groundwater basin thatrecharges the Devils River watershed may extend beyond theboundary of the surface watershed if the average recharge rate isless than 23 mm/yr. Additional assessment is needed to reducethe uncertainty in the estimates for recharge and the baseflow fraction to ascertain the full extent of the Devils River groundwaterbasin.

to account for the amount of recharge water discharged via theDevils River. If recharge is estimated at 20 mm/yr (area! averageof recharge calculated using basin-wide recharge estimates byGreen and Bertetti, 2010), then 12,121 km2 of watershed isrequired to account for the amount of water discharged via the long-term average-flow measurements for rivers that dischargeDevils River. Because the area of the Devils River watershed basin to Amistad Reservoir were calculated by Green and Bertetti (2010)is 10,260 1cm2, this suggests that 15% of the water discharged by using data from the International Boundary and Water Commisthe Devils River to the Amistad Reservoir is sourced from outside sion website (http://www.ibwc.state.gov/Water Data/histof the watershed basin. An average recharge rate of approximately flol.htm). The surface water inputs to Amistad Reservoir are23 mm/yr, however is consistent with a watershed area of 1322 Mm3/yr from the upper Rio Grande, 240 Mm°/yr from the

— Devils/Sycamore Basin

— Streams/Rivers

Wells• Wells <100 (gpml

o Wells 100-499 (gpm)

O Wells 500-999 )gpm)

• Wells 1000+ (9pm)

GainlLossA GainLoss2Ol3

A TCQ Gain Loss 2006

Hydraulic Conductivity

1 5 Irn/day)

15(rnldas)

45 (rn/day)

A0 5 10 20 30 40

Fig. 72. Map of Devils River watershed basin with reRned hydraulic conductivity assignments.

R.T Green et al/Journal of Hydrology 517 (2014) 284—297 295

Pecos River, 325 Mm3/yr from the Devils River, and an estimated125 Mm3/yr from Goodenough Spring now located within AmistadReservoir (Brune, 1975). The total of this input to AmistadReservoir (i.e., 2012 Mm3/yr) compares well with the measureddischarge of 2049 Mm3/yr from Amistad Reservoir. This self-consistent, water-budget analysis indicates that recharge toAmistad Reservoir occurs essentially as surface flow and that thereis negligible inflow to the reservoir as river-channel underfiow orinterformational flow. This suggests that most of the underfiowin the Devils River channel has discharged from the preferentialflow paths to surface flow upstream of the Pafford Crossing rivergauge.

7. Groundwater conveyance in a semi-arid karst terrain

A refined conceptualization of groundwater conveyance in asemi-arid karst terrain is proposed based on fundamental processes of dissolution and using surrogate data for aquifer hydrauliccapacity. Using evidence that indicates a strong correlationbetween aquifer permeability and proximity to higher-order riverchannels, pre-existing representation of the carbonate aquifer’shydraulic properties of the Devils River watershed (Anaya andjones, 2004, 2009; Hutchison et al., 2011) is reinterpreted. Therefined conceptualization of the permeability architecture of thekarst aquifer is proposed in which high-capacity preferential flowpathways coincide with higher order river and stream channels.

Gradational hydraulic property values are assigned to thesepreferential flow paths in the Edwards-Trinity Aquifer in the DevilsRiver watershed based on well pumping capacity. Stream and riverchannels with wells that have capacity greater than 1890 L/min areassigned a hydraulic conductivity of 45 rn/day. Stream and riverchannels with wells that have capacity in the range of 378 L/rninto 1889 C/mm are assigned a hydraulic conductivity of 15 rn/day.All river valleys with enhanced hydraulic conductivity have widthsof 5 1cm, consistent with the correlation distance estimated in thewell capacity/proximity to river assessment. lnterstream areasare assigned a hydraulic conductivity of 1.5 rn/day, a value that isa factor of 30 less than the hydraulic conductivities assigned tothe highest capacity river channels. This relative difference inhydraulic conductivity is comparable to the difference in wellcapacity between wells in the higher order river channels (i.e.,3785 L/rnin) and wells in the interstream areas (i.e., <115 I/mm).A rnap with the refined permeability assignments is presented inFig. 12.

A significant source of uncertainty in these recharge estimatesis determination of the extent of the Devils River watershed basin.Differences between the extents of surface water and groundwaterbasins can be significant in karst systems (Maréchal et al., 2008).Consistent with this generalization, the extent of the Devils Rivergroundwater basin is not well defined and may differ from theextent of the surface watershed. Groundwater rnodeling can beused to estimate recharge in lcarst aquifers (Dorfliger et al., 2009;Fleury et al., 2009), however, successful completion of a rnodelwould be challenging given that both recharge and basin size arepoorly constrained. Nonetheless, recharge, and basin extent, andthe flow dynarnics of the karst aquifer can be refined using a permeability architecture of preferential flow paths coincident withhigher-order river channels. This new frarneworlc would replaceone in which the permeability architecture is based primarily ongeologic mapping. The refined conceptualization is fundamentallyconsistent with (i) lcarst development in carbonate rocks, (ii) structural evolution of the Edwards Plateau, and (iii) the requirementthat the groundwater regime of the Devils River watershed has sufficient capacity to convey sufficient quantities of water at therequired rates across the full extent of the watershed.

8. Conclusions

An efficient conveyance system for groundwater is shown tohave formed in a karst carbonate watershed located in a semi-aridenvironment. This conveyance systern comprises preferential flowpathways that developed coincident with river channels whoselocations appear to date to the early days of regional uplift andexhumation of the carbonate formations. A strong correlationbetween wells with high pumping capacity and proximity tohigher-order river channels (i.e., within 2.5 1cm) was used as evidence of preferential flow pathway presence. The principle factorsthat contributed to development of the preferential flow paths arethe presence of a limestone-rich formation and recharge that hasbeen geomorphologicatly focused toward river channels. A secondary factor that may have contributed to the development of thesepreferential flow paths is the relatively large hydraulic gradient(i.e., in excess of 0.001) enhanced as the Edwards Plateau wasexhumed.

flow measurements in the Devils River measured under relatively high- and low-flow conditions supports the hypothesis thatthe river is gaining in downstream reaches at a rate that exceedsthe added size of the watershed. This characteristic leads to perennial river flow being restricted to only the lower reach of the river.Lastly, water-budget analysis of the Devils River watershedsupports the interpretation that essentially all of the recharge toArnistad Reservoir that is derived from the Devils River watershedis contributed as surface flow from the river and that there is minimal underfiow or cross-formational flow from the watershed atthe point the watershed abuts the reservoir. Recognition of thesepreferential pathways in proximity to river channels provides abasis to determine where high capacity wells are lilcely (and unliIcely) and suggests that groundwater flow within the watershed isrelatively rapid, consistent with flow rates representative of Icarsticaquifers (Worthington, 2007). This understanding provides a basisfor better informed decisions regarding water-resources management in a semi-arid environment.

The Devils River watershed basin in the Edwards-Trinity Aquifer system in south-central Texas was selected to evaluate thisinterpretation and conceptualization. Although the climate of theDevils River watershed is semi-arid, the watershed is the sourcefor significant water resources that contribute to the Rio Grande.The Devils River watershed basin is representative of a broad classof karst carbonate aquifers in semi-arid environments worldwide.Accordingly, groundwater conveyance mechanisms of importancein the Devils River watershed basin may help characterize similarkarst aquifers in other semi-arid environments that provide significant water resources.

Acknowledgments

Support for this work was provided by the Coypu Foundation.Comments by Gary Walter, David Ferrill, and three anonymousreviewers improved the manuscript and are greatly appreciated.The authors are thanlcful tojeffrey Bennett, Kevin Urbanczylc, Marcus Gary, and their students for the conduct of the 2013 flow survey of the Devils River.

References

Abbott, P.C., 1975. On the hydrology of the Edwards Limestone, south-central Texas.J. Hyclrol. 24, 251 269.

Allen, 0.)., Brewerton, L.J., coleby, CM., Gibbs, BR., Lewis, M.A., Wagstaff, 5).,Williams, AT., 1997. The physical properties of major aquifers in England andWales. British Geological Survey Technical, Report WD/97/34. p. 312.

Anaya, R., Jones. I., 2004. Groundwater Availability Model for the Edwards-Trinity(Plateau) and cenozoic Pecos Alluvium Aquifer Systems, Texas. Texas WaterDevelopment Board. Austin, TX. p. 208.

296 RI. Green or aL/Jounial of Hydrology 517 (2014) 284—297

Anaya, R., Jones, I., 2009. Groundwater Availability Model for the Edwards-Trinity(Plateau) and Pecos Valley Aquifers of Texas. Texas Water Development Board.Austin, TX. Report 373. p. 103.

Aquilina. L, Ladouche. B., Doerfliger, N.. 2005. Recharge processes in karstic systemsinvestigated through the correlation of chemical and isotopic composition ofrain and spring waters. AppI. Geochem. 20, 2 189—2206.

Ardis, A.F., Barker, R.A., 1993. Historical Saturated Thickness of the Edwards-TrinityAquifer System and Selected Contiguous Hydraulically Connected Units, West-Central Texas: U.S. Geological Survey Water-Resources Investigation, Report92-4125.2 Plates.

Arnold, J.G., Allen. P.M.. 1999. Automated methods for estimating baseflow andground water recharge from streamfiow records. J. Am. Water Resour. Assoc. 35(2). 411-424.

Arnold, ].G., Allen. P.M.. Muttiah. R., Bernhardt, G., 1995. Automated base flowseparation and recession analysis techniques. Ground Water 33 (6), 1010—1018.

Bakalowicz. M., 2005. Karst groundwater: a challenge for new resources. Hydrogeol.J. 13, 148—160.

Barker, RA, Ardis, A.F., 1992. Configuration of the Base of the Edwards-TrinityAquifer System and Hydrogeology of the Underlying Pre-Cretaceous Rocks,West-Central Texas: U.S. Geological Survey Water Resources Investigation,Report 91-4071, p. 25.

Barker, R.A., Ardis, A.F.. 1996. Hydrogeologic Framework of the Edwards-TrinityAquifer System, West-Central Texas: U.S. Geological Survey Professional Paper1421-B. p. 61. with plates.

Barker, R.A., Bush. P.W.. Baker, Jr., tI., 1994. Geologic History and HydrogeologicSetting of the Edwards-Trinity Aquifer System. West-Central Texas: U.S.Geological Survey Water Resources Investigation. Report 94-4039. p. 50.

Bauer, S., Leidl, R., Sauter. M., 2005. Modeling the influent of epikarsi evolution onkarst aquifer genesis: a time-variant recharge boundary condition for jointkarst-epikarst development. Water Resour. Res. 41. W09416. http://dx.doi.org/I0.102940045VR00332 I.

Brune. G.. 1975. Major and Historical Springs of Texas. Texas. Texas WaterDevelopment Board. Report 189. p. 94.

Bush, P.W., Ardis. A.F., Wynn, K.H., 1993. Historical Potentiometric Surface of theEdwards-Trinity Aquifer System and Contiguous Hydraulically Connected Units,West-Central Texas: U.S. Geological Survey Water-Resources Investigations,Report 92-4055. 3 sheets.

Connorton, B]., Reed, RN.. 1978. A numerical model for the prediction of long termwell yield in an unconfined chalk aquifer. Q. JI. tngng. Ceol. 11, 127—138.

Doctor, D.H., Alexander Jr., E.C., Petrif. M.. Kogoviek.].. Urbanc,J., Lojen, S., Stichler,W., 2006. Quantification of karst aquifer discharge components during stormevents through end-member mixing analysis using natural chemistry andstable isotopes as tracers. Hydrogeol. J. 14 (1). http://dx.doi.org/10.1007/sIOO4O-006-0031-6, pp. 1,171—1,191.

Dorfiiger, N., Fletiiy. P., Ladouche. B., 2009. Inverse modeling approach to allogenickarst system characterization. Ground Water 47, 414—426.

Dreybrodt, W., Gabrovsek, F., 2003. Basic processes and mechanisms governing theevolution of karst speleogenesis and evolution of karst aquifers. SpeleogenesisEvol. Karst Aquifers 1(1), 115—154.

Dreybrodt, W., Gabrov8ek, F., Romanov, D., 2005. Processes of speleogenesis: amodeling approach. Karst Research Institute at ZRC SAZU, Postojna — Ljubljana, p.376.

El-Hakim, M., Bakalowicz, M., 2007. Significance and origin of very large regulatingpower of some karst aquifers in the Middle East. Implication on karst aquiferclassification. J. Hydrol. 333, 329—339.

Fisher, W.L., 1977. Geologic Atlas of Texas: Del Rio Sheet. Scale: 1:125,000. Bureauof Economic Geology, The University of Texas at Austin. Austin, Texas.

Fisher, W.L, 1981. Geologic Atlas of Texas: Sonora Sheet. Scale: 1:125,000. Bureauof Economic Geology, The University of Texas at Austin. Austin, Texas.

Fleury, P., Ladouche, B., Conroux, Y., Jotirde, H., Dorfliger. N.. 2009. Modelling thehydrologic functions of a karst aquifer under active water management — theLez spring.]. Hydrol. 365. 235—243.

Ford, D., Williams, P., 1989. Karst Geomorphology and Hydrology: Unwin Hyman,Winchester, Massachusetts, p. 320.

Green, R.T., Bertetti, F.P., 2010. Investigating the Water Resources of the WesternEdwards-Trinity Aquifer, Texas. Abstracts for the 2010 Geological Society ofAmerica Annual Conference.

Green, RI., Bertetti, f.P., 2012. Groundwater resource management in sub-humidand semi-arid environments. Presented at the National GroundwaterAssociation Emerging Issues in Groundwater Conference. San Antonio, Texas,February 26—27, 2012.

Green, R., Bertetti, F., Hernandez, M., 2012. Recharge variability in semi-aridclimates. Nat Edu Knowl 3 (3), 10.

Gregory. K.J., 1976. Drainage networks and climate. Chapter 10 in Geomorphologyand Climate. Derbyshire, E. (Ed.). John Wiley, London.

Hartmann, A.. Lange,J., Weiler, M., Arbel. Y.. Greenbaum, N., 2012. A new approachto model the spatial and temporal variability of recharge to karst aquifers.Hydrol. Earth Syst. Sci. 16, 2219—2231.

Hartmann, A,, Barberd, J.A., Lange, J., Andreo. B., Weiler, M.. 2013. Progress in thehydrologic simulation of time variant rerli,iige areas of karst systems —

exemplified at a karst spring in Southern Spain. Adv. Water Resour. S-I, 149—160.Hughes, AG., Mansour. MM., 2005. Recharge modelling for the West Bank aquifers.

British Geological Survey Commissioned Report No. CR/05/087N. BritishGeological Survey. In: International Boundary and Water Commission. 2005.Water Bulletin Number 75: Flow of the Rio Grande and Related Data; FromElephant Butte Dam, New Mexico to the Gulf of Mexico.

Hutchison, W.R., Jones, IC., Anaya, R., 2011. Update of the Groundwater AvailabilityModel for the Edwards-Trinity (Plateau) and Pecos Valley Aquifers of Texas.Texas Water Development Board, p. 61.

Klimchouk, A.B., Ford, D.C., 2000a. Types of Isarst and evolution of hydrogeologicsetting. In: Speleogenesis Evolution of Karst Aquifers. National SpeleologicalSociety, Huntsville, AL, United States.

I<Iimchouk, A.B., Ford, D.C., 2000b. Lithological and structural controls ofdissolutional cave development. In: Speleogenesis Evolution of I<arst Aquifers.National Speleological Society, Huntsville, AL, United States.

I<uniansky, EL., Holligan, K.Q., 1994. Simulations of Flow in the Edwards-TrinityAquifer System and Contiguous Hydraulically Connected Units, West-CentralTexas. US. Geological Survey. Water-Resources Investigations, Report 93-4039,p. 40.

Latiman, L.H., Parizek, R.R.. 1964. Relationship between fracture traces and theoccurrence of grocind water in carbonate rocks. J. Hydrol. 2,73—91.

Mabee, SB., Hardcastle, K.C., Wise, D.U., 1994. A method of collecting and analysinglineaments for regional-scale fractured-bedrock aquifer studies. Groundwater32 (6), 884—894.

Mabee, SB., Curry. P.]., Hardcastle. K.C.. 2002. Correlation of Iineaments to groundwater inflows in a bedrock tunnel. Ground Water 40(1). 37-43.

MacDonald. AM.. AlIen, Di.. 2001. Aquifer properties of the Chalk of England. Quart.J. Eng. Geol. Hydrogeol. 34, 371—384.

Magowe, M., 1999. Relationship berween Iineaments and ground water occtirrencesin western Botswana. Ground Water 37 (2). 282—286.

Mardchal, ].C., Ladouche, B., Dorfilger, N., 2008. Karst flash flooding in aMediterranean karst, the example of Fontaine do Nimes. Eng. Geol. 99, 138—146.

Martinez-Santos, P., Andreu, J.M., 2010. Lumped and distributed approaches tomodel natural recharge in semiarid karst aquifers. J. Hydrol. 388, 189—398.

Mocochain. L. Audra, P.. Clauzon. G.. Bellier, 0.. Bigot. ].-Y.. Parize. 0., Monteil, P.,2009. The effect of river dynamics induced by the Messinian Salinity Crisis onkarst landscape and caves: example of the Lower Arddche river (mid Rhónevalley). Geomorphology 106, 46-61.

Moore, R.B.. Schwarz, G.E., Clark, Jr., S.F., Walsh, G.J., Degnan, JR., 2002. factorsrelated to well yield in the fractured-bedrock aquifer of New Hampshire. UnitedStates Geologic Survey Professional Paper, vol. 1660. p. 18.

Morel. t.H., 1980. The use of a numerical model in the management of the Chalkaquifer in the Upper Thames Basin. Q. ]l. Engng. Geol. 13, 153—165.

Moon, 5., 2004. Using streams and faults as lineamenrs to delineate aquifercharacteristics. MS. Thesis. Oklahoma State University, p. 63.

Mouri, S., Halihan, I., 2007. Averaging Hydraulic Conductivity in HeterogeneousFractures and Layered Aquifers. In: Icrasny, J., Sharp. Jr., J.M. (Eds.), SelectedPapers on Hydrogeology, International Association of Hydrogeologists, A.A.Balkema, Rotterdam.

Palmer, AN., 1991. Origin and morphology of limestone caves. Geol. Soc. Am. Bull.103, 1—21.

Parizek, R.R., 1975. On the Nature and Significance of Fracture Traces andLineaments in Carbonate and Other Terranes. In: Karst Hydrology and WaterResources Proceedings of the U.S-Yugoslavian Symposium, pp. 47—108.

Rashed, K.A., 2012. Assessing degree of karstification: a new method of classifyingkarst aquifers. In: Sixteenth International Water Technology Conference, IWTC16 2012, Istanbul, Turkey.

Reeves, RD., Small. TA., 1973. Groundwater Resources of Val Verde County, Texas.Report 172. Texas Water Development Board, p. 143.

Rodriguez-lturbe, I., Escobar, L.A., 1982. The dependence of drainage density onclimate and geomorphology. Hydrol. Sci. J. 27 (2), 129—137. http://dx.doi.org/10.1080/02626668209491095.

Romanov, D., Gabrovsek, F., Dreybrodt, W., 2003. The impact of hydrochemicalboundary conditions on the evolution of limestone karst aquifers.]. Hydrol. 276,240—253.

Rose, P.R., 1972. Edwards Group surface and subsurface, Central Texas. Bureau ofEconomic Geology, The University of Texas at Austin. Austin, Texas, p. 198. +

attachments.Sander, P., Minor, TB., Chesley, MM., 1996. Ground-water exploration based on

lineament analysis and reproducibility tests. Ground Water 35 (5), 88—894.Scanlon, BR., Mace, RE., Barrett, ME., Smith, B.. 2003. Can we simulate regional

groundwater flow in a karst system using equivalent porous media models?Case study, Barton Springs Edwards aquifei. USA J. Hydrol. 276. 137—158.

Schindel, G.M., Quinlan, J.F., Davies. G., Ray, J.A.. 1996. Guidelines for WellheadProtection and Springhead Protection Area Delineation in Carbonate Rocks, EPA904-b-97-003.

Sharpe, W.f.. Parizek, R.R., 1979. Ground water sources located by fracture tracetechnique. Water Sewage Works 126 (5), 38—40.

Siddiqui. S.H., Parizek, R.R., 1971. Hydrologic factors influencing well-yields infolded and fatilted carbonated rocks in central Pennsylvania. Water Resour. Res.7(5). 1295—1312.

Stricklin, Jr.. FL, Smith, Cl., Lozo, FE... 1971. Stratigraphy of Lower CretaceousTrinity Deposits of Central Texas. Bureau of Economic Geology, The Universityof Texas at Austin. Austin, Texas, Report of Investigations No. 71.

Texas Commission for Environmental Quality, 2006. Devils River Gain/Loss Survey.Surface Water Quality Management. Region 13. Texas Commission forEnvironmental Quality.

Texas Water Development Board (PNDB), 2012a. Groundwater Database. http://www.twdb.stare.tx.us/groundwater/data/gwdbrpt.asp. Accessed July 16, 2012.

Texas Water Development Board (TWDB). 2012b. Submitted Driller’s ReportsDatabase. <http://www.twdb.state.tx.us/groundwater/dara/drillersdb.asp>.Accessed December 5,2012.

RI. Green et al/journal of Hydrology 517 (2014) 284—297 297

United States Geological Survey, 2013. National Hydrography Dataset. http://nhd.usgs.gov/. Accessed on September 23, 2013.

Veni, G., Associates, 1996. Drainage basin delineation and preliminary hydrologicassessment of the Dolan Springs, Val Verde County, Texas. Prepared for TheNature Conservancy of Texas, vol. 1.

White, W.B., 1999. Conceptual Models for ICarstic Aquifers. In: Palmer, AN., Palmer,MV., Sasowsky, ID., (Eds.). l<arst Modeling: Special Publication 5, The KarstWaters Institute, Charles Town. West Virginia. pp. 11—16.

White, W,B,, 2006. fifty Years of Karst Hydrology and Hydrogeology: 1953—2003.In: Harmon, R.S., Wicks, C. fEds.), Perspectives on karst geomorphology,hydrology, and geochemistry: a tribute volume to Derek C. Ford and WilliamB, White. Geological Society of America Special Paper 404, pp. 139—152.

White, \,V.B., White. EL., 2001. Conduit fragment,ition. c,sve patterns, and thelocalization of karst ground water basins: the Appalachians as a test case. Theor.AppI. Karstology 13—14, 9—24,

Wilcox, B.P., Wilding, L.P., Woodruffjr., CM., 2007. Soil and topographic controls onrunoff generation from stepped landforms in the Edwards Plateau of CentralTexas. Geophy. Res. Let. 34, L24S24. http://dx.doi.orgjlO.l029/2007GL030860.

WoodruftJr., CM.. 1974. Evidence for stream piracy along the Balcones escarpment,central Texas. Geol. Soc. Am., Abstr Prog. 6(7), 1010.

Woodruff Jr., CM., 1977, Stream piracy near the Balcones fault zone, Cent. Tex. J.Geol. 85 (4), 483-490.

Woodruff Jr., CM,, Abbott, P.L, 1979. Drainage-basin evolution and aquiferdevelopment in a karstic limestone terrain, South-Central Texas. USA EarthSurf. Process, 4, 319—334.

Woodruff Jr., CM.. Abbott. P.L,. 1986. Stream piracy and evolution of the EdwardsAquifer along the Balcones Escarpment, Central Texas, In: Abbott, P.L,Woodruff, CM., Jr. fEds.), The Balcones Escarpment. Geological Society ofAmerica, Central Texas, pp. 77—90.

Woodruff]r., CM., Wilding, L.P.. 2008. Bedrock, soils. and hillslope hydrology in theCentral Texas Hill Cotintry, USA: implications on environmental management ina carbonate-rock terrain. Environ. Geol. 55 (3), 605—618.

Worthington, S.R.H.. 2007. Ground-water residence times in unconBned carbonateaquifers. J. Cave Karst Stud. 69(1), 94—102.

Worthington, S.R.H.. Ford, D.C., 2009. Self-organized permeability in carbonateaquifers. Ground Water 47 (3), 326—336,

Worthingtois, S.R.H.. ford. D.C.. Beddows, PA.. 2000. Porosity and PermeabilityEnhancement in Unconhned Carbonate Aquifers as a Result of Solution inSpeleogenesis. In: Klimchouk, A., Ford, D.C., Palmer, AN., Dreybrodt, W. tEds.),Evolution of Karst Aquifers. National Speleological Society Inc. Huntsville. AL.USA, pp. 463-472.